4 * Copyright (C) 1991, 1992 Linus Torvalds
7 #include <linux/export.h>
9 #include <linux/utsname.h>
10 #include <linux/mman.h>
11 #include <linux/reboot.h>
12 #include <linux/prctl.h>
13 #include <linux/highuid.h>
15 #include <linux/kmod.h>
16 #include <linux/perf_event.h>
17 #include <linux/resource.h>
18 #include <linux/kernel.h>
19 #include <linux/kexec.h>
20 #include <linux/workqueue.h>
21 #include <linux/capability.h>
22 #include <linux/device.h>
23 #include <linux/key.h>
24 #include <linux/times.h>
25 #include <linux/posix-timers.h>
26 #include <linux/security.h>
27 #include <linux/dcookies.h>
28 #include <linux/suspend.h>
29 #include <linux/tty.h>
30 #include <linux/signal.h>
31 #include <linux/cn_proc.h>
32 #include <linux/getcpu.h>
33 #include <linux/task_io_accounting_ops.h>
34 #include <linux/seccomp.h>
35 #include <linux/cpu.h>
36 #include <linux/personality.h>
37 #include <linux/ptrace.h>
38 #include <linux/fs_struct.h>
39 #include <linux/file.h>
40 #include <linux/mount.h>
41 #include <linux/gfp.h>
42 #include <linux/syscore_ops.h>
43 #include <linux/version.h>
44 #include <linux/ctype.h>
46 #include <linux/mempolicy.h>
47 #include <linux/sched.h>
49 #include <linux/compat.h>
50 #include <linux/syscalls.h>
51 #include <linux/kprobes.h>
52 #include <linux/user_namespace.h>
53 #include <linux/binfmts.h>
55 #include <linux/sched.h>
56 #include <linux/rcupdate.h>
57 #include <linux/uidgid.h>
58 #include <linux/cred.h>
60 #include <linux/kmsg_dump.h>
61 /* Move somewhere else to avoid recompiling? */
62 #include <generated/utsrelease.h>
64 #include <asm/uaccess.h>
66 #include <asm/unistd.h>
68 #ifdef CONFIG_MT_PRIO_TRACER
69 # include <linux/prio_tracer.h>
72 #ifndef SET_UNALIGN_CTL
73 # define SET_UNALIGN_CTL(a,b) (-EINVAL)
75 #ifndef GET_UNALIGN_CTL
76 # define GET_UNALIGN_CTL(a,b) (-EINVAL)
79 # define SET_FPEMU_CTL(a,b) (-EINVAL)
82 # define GET_FPEMU_CTL(a,b) (-EINVAL)
85 # define SET_FPEXC_CTL(a,b) (-EINVAL)
88 # define GET_FPEXC_CTL(a,b) (-EINVAL)
91 # define GET_ENDIAN(a,b) (-EINVAL)
94 # define SET_ENDIAN(a,b) (-EINVAL)
97 # define GET_TSC_CTL(a) (-EINVAL)
100 # define SET_TSC_CTL(a) (-EINVAL)
104 * this is where the system-wide overflow UID and GID are defined, for
105 * architectures that now have 32-bit UID/GID but didn't in the past
108 int overflowuid
= DEFAULT_OVERFLOWUID
;
109 int overflowgid
= DEFAULT_OVERFLOWGID
;
111 EXPORT_SYMBOL(overflowuid
);
112 EXPORT_SYMBOL(overflowgid
);
115 * the same as above, but for filesystems which can only store a 16-bit
116 * UID and GID. as such, this is needed on all architectures
119 int fs_overflowuid
= DEFAULT_FS_OVERFLOWUID
;
120 int fs_overflowgid
= DEFAULT_FS_OVERFLOWUID
;
122 EXPORT_SYMBOL(fs_overflowuid
);
123 EXPORT_SYMBOL(fs_overflowgid
);
126 * this indicates whether you can reboot with ctrl-alt-del: the default is yes
131 EXPORT_SYMBOL(cad_pid
);
134 * If set, this is used for preparing the system to power off.
137 void (*pm_power_off_prepare
)(void);
140 * Returns true if current's euid is same as p's uid or euid,
141 * or has CAP_SYS_NICE to p's user_ns.
143 * Called with rcu_read_lock, creds are safe
145 static bool set_one_prio_perm(struct task_struct
*p
)
147 const struct cred
*cred
= current_cred(), *pcred
= __task_cred(p
);
149 if (uid_eq(pcred
->uid
, cred
->euid
) ||
150 uid_eq(pcred
->euid
, cred
->euid
))
152 if (ns_capable(pcred
->user_ns
, CAP_SYS_NICE
))
158 * set the priority of a task
159 * - the caller must hold the RCU read lock
161 static int set_one_prio(struct task_struct
*p
, int niceval
, int error
)
165 if (!set_one_prio_perm(p
)) {
169 if (niceval
< task_nice(p
) && !can_nice(p
, niceval
)) {
173 no_nice
= security_task_setnice(p
, niceval
);
180 #ifdef CONFIG_MT_PRIO_TRACER
181 set_user_nice_syscall(p
, niceval
);
183 set_user_nice(p
, niceval
);
189 SYSCALL_DEFINE3(setpriority
, int, which
, int, who
, int, niceval
)
191 struct task_struct
*g
, *p
;
192 struct user_struct
*user
;
193 const struct cred
*cred
= current_cred();
198 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
201 /* normalize: avoid signed division (rounding problems) */
209 read_lock(&tasklist_lock
);
213 p
= find_task_by_vpid(who
);
217 error
= set_one_prio(p
, niceval
, error
);
221 pgrp
= find_vpid(who
);
223 pgrp
= task_pgrp(current
);
224 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
225 error
= set_one_prio(p
, niceval
, error
);
226 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
229 uid
= make_kuid(cred
->user_ns
, who
);
233 else if (!uid_eq(uid
, cred
->uid
) &&
234 !(user
= find_user(uid
)))
235 goto out_unlock
; /* No processes for this user */
237 do_each_thread(g
, p
) {
238 if (uid_eq(task_uid(p
), uid
))
239 error
= set_one_prio(p
, niceval
, error
);
240 } while_each_thread(g
, p
);
241 if (!uid_eq(uid
, cred
->uid
))
242 free_uid(user
); /* For find_user() */
246 read_unlock(&tasklist_lock
);
253 * Ugh. To avoid negative return values, "getpriority()" will
254 * not return the normal nice-value, but a negated value that
255 * has been offset by 20 (ie it returns 40..1 instead of -20..19)
256 * to stay compatible.
258 SYSCALL_DEFINE2(getpriority
, int, which
, int, who
)
260 struct task_struct
*g
, *p
;
261 struct user_struct
*user
;
262 const struct cred
*cred
= current_cred();
263 long niceval
, retval
= -ESRCH
;
267 if (which
> PRIO_USER
|| which
< PRIO_PROCESS
)
271 read_lock(&tasklist_lock
);
275 p
= find_task_by_vpid(who
);
279 niceval
= 20 - task_nice(p
);
280 if (niceval
> retval
)
286 pgrp
= find_vpid(who
);
288 pgrp
= task_pgrp(current
);
289 do_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
) {
290 niceval
= 20 - task_nice(p
);
291 if (niceval
> retval
)
293 } while_each_pid_thread(pgrp
, PIDTYPE_PGID
, p
);
296 uid
= make_kuid(cred
->user_ns
, who
);
300 else if (!uid_eq(uid
, cred
->uid
) &&
301 !(user
= find_user(uid
)))
302 goto out_unlock
; /* No processes for this user */
304 do_each_thread(g
, p
) {
305 if (uid_eq(task_uid(p
), uid
)) {
306 niceval
= 20 - task_nice(p
);
307 if (niceval
> retval
)
310 } while_each_thread(g
, p
);
311 if (!uid_eq(uid
, cred
->uid
))
312 free_uid(user
); /* for find_user() */
316 read_unlock(&tasklist_lock
);
323 * emergency_restart - reboot the system
325 * Without shutting down any hardware or taking any locks
326 * reboot the system. This is called when we know we are in
327 * trouble so this is our best effort to reboot. This is
328 * safe to call in interrupt context.
330 void emergency_restart(void)
332 kmsg_dump(KMSG_DUMP_EMERG
);
333 machine_emergency_restart();
335 EXPORT_SYMBOL_GPL(emergency_restart
);
337 void kernel_restart_prepare(char *cmd
)
339 blocking_notifier_call_chain(&reboot_notifier_list
, SYS_RESTART
, cmd
);
340 system_state
= SYSTEM_RESTART
;
341 usermodehelper_disable();
346 * register_reboot_notifier - Register function to be called at reboot time
347 * @nb: Info about notifier function to be called
349 * Registers a function with the list of functions
350 * to be called at reboot time.
352 * Currently always returns zero, as blocking_notifier_chain_register()
353 * always returns zero.
355 int register_reboot_notifier(struct notifier_block
*nb
)
357 return blocking_notifier_chain_register(&reboot_notifier_list
, nb
);
359 EXPORT_SYMBOL(register_reboot_notifier
);
362 * unregister_reboot_notifier - Unregister previously registered reboot notifier
363 * @nb: Hook to be unregistered
365 * Unregisters a previously registered reboot
368 * Returns zero on success, or %-ENOENT on failure.
370 int unregister_reboot_notifier(struct notifier_block
*nb
)
372 return blocking_notifier_chain_unregister(&reboot_notifier_list
, nb
);
374 EXPORT_SYMBOL(unregister_reboot_notifier
);
376 /* Add backwards compatibility for stable trees. */
377 #ifndef PF_NO_SETAFFINITY
378 #define PF_NO_SETAFFINITY PF_THREAD_BOUND
381 static void migrate_to_reboot_cpu(void)
383 /* The boot cpu is always logical cpu 0 */
386 cpu_hotplug_disable();
388 /* Make certain the cpu I'm about to reboot on is online */
389 if (!cpu_online(cpu
))
390 cpu
= cpumask_first(cpu_online_mask
);
392 /* Prevent races with other tasks migrating this task */
393 current
->flags
|= PF_NO_SETAFFINITY
;
395 /* Make certain I only run on the appropriate processor */
396 set_cpus_allowed_ptr(current
, cpumask_of(cpu
));
400 * kernel_restart - reboot the system
401 * @cmd: pointer to buffer containing command to execute for restart
404 * Shutdown everything and perform a clean reboot.
405 * This is not safe to call in interrupt context.
407 void kernel_restart(char *cmd
)
409 kernel_restart_prepare(cmd
);
410 migrate_to_reboot_cpu();
413 printk(KERN_EMERG
"Restarting system.\n");
415 printk(KERN_EMERG
"Restarting system with command '%s'.\n", cmd
);
416 kmsg_dump(KMSG_DUMP_RESTART
);
417 machine_restart(cmd
);
419 EXPORT_SYMBOL_GPL(kernel_restart
);
421 static void kernel_shutdown_prepare(enum system_states state
)
423 blocking_notifier_call_chain(&reboot_notifier_list
,
424 (state
== SYSTEM_HALT
)?SYS_HALT
:SYS_POWER_OFF
, NULL
);
425 system_state
= state
;
426 usermodehelper_disable();
430 * kernel_halt - halt the system
432 * Shutdown everything and perform a clean system halt.
434 void kernel_halt(void)
436 kernel_shutdown_prepare(SYSTEM_HALT
);
437 migrate_to_reboot_cpu();
439 printk(KERN_EMERG
"System halted.\n");
440 kmsg_dump(KMSG_DUMP_HALT
);
444 EXPORT_SYMBOL_GPL(kernel_halt
);
447 * kernel_power_off - power_off the system
449 * Shutdown everything and perform a clean system power_off.
451 void kernel_power_off(void)
453 kernel_shutdown_prepare(SYSTEM_POWER_OFF
);
454 if (pm_power_off_prepare
)
455 pm_power_off_prepare();
456 migrate_to_reboot_cpu();
458 printk(KERN_EMERG
"Power down.\n");
459 kmsg_dump(KMSG_DUMP_POWEROFF
);
462 EXPORT_SYMBOL_GPL(kernel_power_off
);
464 static DEFINE_MUTEX(reboot_mutex
);
467 * Reboot system call: for obvious reasons only root may call it,
468 * and even root needs to set up some magic numbers in the registers
469 * so that some mistake won't make this reboot the whole machine.
470 * You can also set the meaning of the ctrl-alt-del-key here.
472 * reboot doesn't sync: do that yourself before calling this.
474 SYSCALL_DEFINE4(reboot
, int, magic1
, int, magic2
, unsigned int, cmd
,
477 struct pid_namespace
*pid_ns
= task_active_pid_ns(current
);
481 /* We only trust the superuser with rebooting the system. */
482 if (!ns_capable(pid_ns
->user_ns
, CAP_SYS_BOOT
))
485 /* For safety, we require "magic" arguments. */
486 if (magic1
!= LINUX_REBOOT_MAGIC1
||
487 (magic2
!= LINUX_REBOOT_MAGIC2
&&
488 magic2
!= LINUX_REBOOT_MAGIC2A
&&
489 magic2
!= LINUX_REBOOT_MAGIC2B
&&
490 magic2
!= LINUX_REBOOT_MAGIC2C
))
494 * If pid namespaces are enabled and the current task is in a child
495 * pid_namespace, the command is handled by reboot_pid_ns() which will
498 ret
= reboot_pid_ns(pid_ns
, cmd
);
502 /* Instead of trying to make the power_off code look like
503 * halt when pm_power_off is not set do it the easy way.
505 if ((cmd
== LINUX_REBOOT_CMD_POWER_OFF
) && !pm_power_off
)
506 cmd
= LINUX_REBOOT_CMD_HALT
;
508 mutex_lock(&reboot_mutex
);
510 case LINUX_REBOOT_CMD_RESTART
:
511 kernel_restart(NULL
);
514 case LINUX_REBOOT_CMD_CAD_ON
:
518 case LINUX_REBOOT_CMD_CAD_OFF
:
522 case LINUX_REBOOT_CMD_HALT
:
525 panic("cannot halt");
527 case LINUX_REBOOT_CMD_POWER_OFF
:
532 case LINUX_REBOOT_CMD_RESTART2
:
533 if (strncpy_from_user(&buffer
[0], arg
, sizeof(buffer
) - 1) < 0) {
537 buffer
[sizeof(buffer
) - 1] = '\0';
539 kernel_restart(buffer
);
543 case LINUX_REBOOT_CMD_KEXEC
:
544 ret
= kernel_kexec();
548 #ifdef CONFIG_HIBERNATION
549 case LINUX_REBOOT_CMD_SW_SUSPEND
:
558 mutex_unlock(&reboot_mutex
);
562 static void deferred_cad(struct work_struct
*dummy
)
564 kernel_restart(NULL
);
568 * This function gets called by ctrl-alt-del - ie the keyboard interrupt.
569 * As it's called within an interrupt, it may NOT sync: the only choice
570 * is whether to reboot at once, or just ignore the ctrl-alt-del.
572 void ctrl_alt_del(void)
574 static DECLARE_WORK(cad_work
, deferred_cad
);
577 schedule_work(&cad_work
);
579 kill_cad_pid(SIGINT
, 1);
583 * Unprivileged users may change the real gid to the effective gid
584 * or vice versa. (BSD-style)
586 * If you set the real gid at all, or set the effective gid to a value not
587 * equal to the real gid, then the saved gid is set to the new effective gid.
589 * This makes it possible for a setgid program to completely drop its
590 * privileges, which is often a useful assertion to make when you are doing
591 * a security audit over a program.
593 * The general idea is that a program which uses just setregid() will be
594 * 100% compatible with BSD. A program which uses just setgid() will be
595 * 100% compatible with POSIX with saved IDs.
597 * SMP: There are not races, the GIDs are checked only by filesystem
598 * operations (as far as semantic preservation is concerned).
600 SYSCALL_DEFINE2(setregid
, gid_t
, rgid
, gid_t
, egid
)
602 struct user_namespace
*ns
= current_user_ns();
603 const struct cred
*old
;
608 krgid
= make_kgid(ns
, rgid
);
609 kegid
= make_kgid(ns
, egid
);
611 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
613 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
616 new = prepare_creds();
619 old
= current_cred();
622 if (rgid
!= (gid_t
) -1) {
623 if (gid_eq(old
->gid
, krgid
) ||
624 gid_eq(old
->egid
, krgid
) ||
625 nsown_capable(CAP_SETGID
))
630 if (egid
!= (gid_t
) -1) {
631 if (gid_eq(old
->gid
, kegid
) ||
632 gid_eq(old
->egid
, kegid
) ||
633 gid_eq(old
->sgid
, kegid
) ||
634 nsown_capable(CAP_SETGID
))
640 if (rgid
!= (gid_t
) -1 ||
641 (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
)))
642 new->sgid
= new->egid
;
643 new->fsgid
= new->egid
;
645 return commit_creds(new);
653 * setgid() is implemented like SysV w/ SAVED_IDS
655 * SMP: Same implicit races as above.
657 SYSCALL_DEFINE1(setgid
, gid_t
, gid
)
659 struct user_namespace
*ns
= current_user_ns();
660 const struct cred
*old
;
665 kgid
= make_kgid(ns
, gid
);
666 if (!gid_valid(kgid
))
669 new = prepare_creds();
672 old
= current_cred();
675 if (nsown_capable(CAP_SETGID
))
676 new->gid
= new->egid
= new->sgid
= new->fsgid
= kgid
;
677 else if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->sgid
))
678 new->egid
= new->fsgid
= kgid
;
682 return commit_creds(new);
690 * change the user struct in a credentials set to match the new UID
692 static int set_user(struct cred
*new)
694 struct user_struct
*new_user
;
696 new_user
= alloc_uid(new->uid
);
701 * We don't fail in case of NPROC limit excess here because too many
702 * poorly written programs don't check set*uid() return code, assuming
703 * it never fails if called by root. We may still enforce NPROC limit
704 * for programs doing set*uid()+execve() by harmlessly deferring the
705 * failure to the execve() stage.
707 if (atomic_read(&new_user
->processes
) >= rlimit(RLIMIT_NPROC
) &&
708 new_user
!= INIT_USER
)
709 current
->flags
|= PF_NPROC_EXCEEDED
;
711 current
->flags
&= ~PF_NPROC_EXCEEDED
;
714 new->user
= new_user
;
719 * Unprivileged users may change the real uid to the effective uid
720 * or vice versa. (BSD-style)
722 * If you set the real uid at all, or set the effective uid to a value not
723 * equal to the real uid, then the saved uid is set to the new effective uid.
725 * This makes it possible for a setuid program to completely drop its
726 * privileges, which is often a useful assertion to make when you are doing
727 * a security audit over a program.
729 * The general idea is that a program which uses just setreuid() will be
730 * 100% compatible with BSD. A program which uses just setuid() will be
731 * 100% compatible with POSIX with saved IDs.
733 SYSCALL_DEFINE2(setreuid
, uid_t
, ruid
, uid_t
, euid
)
735 struct user_namespace
*ns
= current_user_ns();
736 const struct cred
*old
;
741 kruid
= make_kuid(ns
, ruid
);
742 keuid
= make_kuid(ns
, euid
);
744 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
746 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
749 new = prepare_creds();
752 old
= current_cred();
755 if (ruid
!= (uid_t
) -1) {
757 if (!uid_eq(old
->uid
, kruid
) &&
758 !uid_eq(old
->euid
, kruid
) &&
759 !nsown_capable(CAP_SETUID
))
763 if (euid
!= (uid_t
) -1) {
765 if (!uid_eq(old
->uid
, keuid
) &&
766 !uid_eq(old
->euid
, keuid
) &&
767 !uid_eq(old
->suid
, keuid
) &&
768 !nsown_capable(CAP_SETUID
))
772 if (!uid_eq(new->uid
, old
->uid
)) {
773 retval
= set_user(new);
777 if (ruid
!= (uid_t
) -1 ||
778 (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
)))
779 new->suid
= new->euid
;
780 new->fsuid
= new->euid
;
782 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RE
);
786 return commit_creds(new);
794 * setuid() is implemented like SysV with SAVED_IDS
796 * Note that SAVED_ID's is deficient in that a setuid root program
797 * like sendmail, for example, cannot set its uid to be a normal
798 * user and then switch back, because if you're root, setuid() sets
799 * the saved uid too. If you don't like this, blame the bright people
800 * in the POSIX committee and/or USG. Note that the BSD-style setreuid()
801 * will allow a root program to temporarily drop privileges and be able to
802 * regain them by swapping the real and effective uid.
804 SYSCALL_DEFINE1(setuid
, uid_t
, uid
)
806 struct user_namespace
*ns
= current_user_ns();
807 const struct cred
*old
;
812 kuid
= make_kuid(ns
, uid
);
813 if (!uid_valid(kuid
))
816 new = prepare_creds();
819 old
= current_cred();
822 if (nsown_capable(CAP_SETUID
)) {
823 new->suid
= new->uid
= kuid
;
824 if (!uid_eq(kuid
, old
->uid
)) {
825 retval
= set_user(new);
829 } else if (!uid_eq(kuid
, old
->uid
) && !uid_eq(kuid
, new->suid
)) {
833 new->fsuid
= new->euid
= kuid
;
835 retval
= security_task_fix_setuid(new, old
, LSM_SETID_ID
);
839 return commit_creds(new);
848 * This function implements a generic ability to update ruid, euid,
849 * and suid. This allows you to implement the 4.4 compatible seteuid().
851 SYSCALL_DEFINE3(setresuid
, uid_t
, ruid
, uid_t
, euid
, uid_t
, suid
)
853 struct user_namespace
*ns
= current_user_ns();
854 const struct cred
*old
;
857 kuid_t kruid
, keuid
, ksuid
;
859 kruid
= make_kuid(ns
, ruid
);
860 keuid
= make_kuid(ns
, euid
);
861 ksuid
= make_kuid(ns
, suid
);
863 if ((ruid
!= (uid_t
) -1) && !uid_valid(kruid
))
866 if ((euid
!= (uid_t
) -1) && !uid_valid(keuid
))
869 if ((suid
!= (uid_t
) -1) && !uid_valid(ksuid
))
872 new = prepare_creds();
876 old
= current_cred();
879 if (!nsown_capable(CAP_SETUID
)) {
880 if (ruid
!= (uid_t
) -1 && !uid_eq(kruid
, old
->uid
) &&
881 !uid_eq(kruid
, old
->euid
) && !uid_eq(kruid
, old
->suid
))
883 if (euid
!= (uid_t
) -1 && !uid_eq(keuid
, old
->uid
) &&
884 !uid_eq(keuid
, old
->euid
) && !uid_eq(keuid
, old
->suid
))
886 if (suid
!= (uid_t
) -1 && !uid_eq(ksuid
, old
->uid
) &&
887 !uid_eq(ksuid
, old
->euid
) && !uid_eq(ksuid
, old
->suid
))
891 if (ruid
!= (uid_t
) -1) {
893 if (!uid_eq(kruid
, old
->uid
)) {
894 retval
= set_user(new);
899 if (euid
!= (uid_t
) -1)
901 if (suid
!= (uid_t
) -1)
903 new->fsuid
= new->euid
;
905 retval
= security_task_fix_setuid(new, old
, LSM_SETID_RES
);
909 return commit_creds(new);
916 SYSCALL_DEFINE3(getresuid
, uid_t __user
*, ruidp
, uid_t __user
*, euidp
, uid_t __user
*, suidp
)
918 const struct cred
*cred
= current_cred();
920 uid_t ruid
, euid
, suid
;
922 ruid
= from_kuid_munged(cred
->user_ns
, cred
->uid
);
923 euid
= from_kuid_munged(cred
->user_ns
, cred
->euid
);
924 suid
= from_kuid_munged(cred
->user_ns
, cred
->suid
);
926 if (!(retval
= put_user(ruid
, ruidp
)) &&
927 !(retval
= put_user(euid
, euidp
)))
928 retval
= put_user(suid
, suidp
);
934 * Same as above, but for rgid, egid, sgid.
936 SYSCALL_DEFINE3(setresgid
, gid_t
, rgid
, gid_t
, egid
, gid_t
, sgid
)
938 struct user_namespace
*ns
= current_user_ns();
939 const struct cred
*old
;
942 kgid_t krgid
, kegid
, ksgid
;
944 krgid
= make_kgid(ns
, rgid
);
945 kegid
= make_kgid(ns
, egid
);
946 ksgid
= make_kgid(ns
, sgid
);
948 if ((rgid
!= (gid_t
) -1) && !gid_valid(krgid
))
950 if ((egid
!= (gid_t
) -1) && !gid_valid(kegid
))
952 if ((sgid
!= (gid_t
) -1) && !gid_valid(ksgid
))
955 new = prepare_creds();
958 old
= current_cred();
961 if (!nsown_capable(CAP_SETGID
)) {
962 if (rgid
!= (gid_t
) -1 && !gid_eq(krgid
, old
->gid
) &&
963 !gid_eq(krgid
, old
->egid
) && !gid_eq(krgid
, old
->sgid
))
965 if (egid
!= (gid_t
) -1 && !gid_eq(kegid
, old
->gid
) &&
966 !gid_eq(kegid
, old
->egid
) && !gid_eq(kegid
, old
->sgid
))
968 if (sgid
!= (gid_t
) -1 && !gid_eq(ksgid
, old
->gid
) &&
969 !gid_eq(ksgid
, old
->egid
) && !gid_eq(ksgid
, old
->sgid
))
973 if (rgid
!= (gid_t
) -1)
975 if (egid
!= (gid_t
) -1)
977 if (sgid
!= (gid_t
) -1)
979 new->fsgid
= new->egid
;
981 return commit_creds(new);
988 SYSCALL_DEFINE3(getresgid
, gid_t __user
*, rgidp
, gid_t __user
*, egidp
, gid_t __user
*, sgidp
)
990 const struct cred
*cred
= current_cred();
992 gid_t rgid
, egid
, sgid
;
994 rgid
= from_kgid_munged(cred
->user_ns
, cred
->gid
);
995 egid
= from_kgid_munged(cred
->user_ns
, cred
->egid
);
996 sgid
= from_kgid_munged(cred
->user_ns
, cred
->sgid
);
998 if (!(retval
= put_user(rgid
, rgidp
)) &&
999 !(retval
= put_user(egid
, egidp
)))
1000 retval
= put_user(sgid
, sgidp
);
1007 * "setfsuid()" sets the fsuid - the uid used for filesystem checks. This
1008 * is used for "access()" and for the NFS daemon (letting nfsd stay at
1009 * whatever uid it wants to). It normally shadows "euid", except when
1010 * explicitly set by setfsuid() or for access..
1012 SYSCALL_DEFINE1(setfsuid
, uid_t
, uid
)
1014 const struct cred
*old
;
1019 old
= current_cred();
1020 old_fsuid
= from_kuid_munged(old
->user_ns
, old
->fsuid
);
1022 kuid
= make_kuid(old
->user_ns
, uid
);
1023 if (!uid_valid(kuid
))
1026 new = prepare_creds();
1030 if (uid_eq(kuid
, old
->uid
) || uid_eq(kuid
, old
->euid
) ||
1031 uid_eq(kuid
, old
->suid
) || uid_eq(kuid
, old
->fsuid
) ||
1032 nsown_capable(CAP_SETUID
)) {
1033 if (!uid_eq(kuid
, old
->fsuid
)) {
1035 if (security_task_fix_setuid(new, old
, LSM_SETID_FS
) == 0)
1049 * Samma pa svenska..
1051 SYSCALL_DEFINE1(setfsgid
, gid_t
, gid
)
1053 const struct cred
*old
;
1058 old
= current_cred();
1059 old_fsgid
= from_kgid_munged(old
->user_ns
, old
->fsgid
);
1061 kgid
= make_kgid(old
->user_ns
, gid
);
1062 if (!gid_valid(kgid
))
1065 new = prepare_creds();
1069 if (gid_eq(kgid
, old
->gid
) || gid_eq(kgid
, old
->egid
) ||
1070 gid_eq(kgid
, old
->sgid
) || gid_eq(kgid
, old
->fsgid
) ||
1071 nsown_capable(CAP_SETGID
)) {
1072 if (!gid_eq(kgid
, old
->fsgid
)) {
1087 * sys_getpid - return the thread group id of the current process
1089 * Note, despite the name, this returns the tgid not the pid. The tgid and
1090 * the pid are identical unless CLONE_THREAD was specified on clone() in
1091 * which case the tgid is the same in all threads of the same group.
1093 * This is SMP safe as current->tgid does not change.
1095 SYSCALL_DEFINE0(getpid
)
1097 return task_tgid_vnr(current
);
1100 /* Thread ID - the internal kernel "pid" */
1101 SYSCALL_DEFINE0(gettid
)
1103 return task_pid_vnr(current
);
1107 * Accessing ->real_parent is not SMP-safe, it could
1108 * change from under us. However, we can use a stale
1109 * value of ->real_parent under rcu_read_lock(), see
1110 * release_task()->call_rcu(delayed_put_task_struct).
1112 SYSCALL_DEFINE0(getppid
)
1117 pid
= task_tgid_vnr(rcu_dereference(current
->real_parent
));
1123 SYSCALL_DEFINE0(getuid
)
1125 /* Only we change this so SMP safe */
1126 return from_kuid_munged(current_user_ns(), current_uid());
1129 SYSCALL_DEFINE0(geteuid
)
1131 /* Only we change this so SMP safe */
1132 return from_kuid_munged(current_user_ns(), current_euid());
1135 SYSCALL_DEFINE0(getgid
)
1137 /* Only we change this so SMP safe */
1138 return from_kgid_munged(current_user_ns(), current_gid());
1141 SYSCALL_DEFINE0(getegid
)
1143 /* Only we change this so SMP safe */
1144 return from_kgid_munged(current_user_ns(), current_egid());
1147 void do_sys_times(struct tms
*tms
)
1149 cputime_t tgutime
, tgstime
, cutime
, cstime
;
1151 spin_lock_irq(¤t
->sighand
->siglock
);
1152 thread_group_cputime_adjusted(current
, &tgutime
, &tgstime
);
1153 cutime
= current
->signal
->cutime
;
1154 cstime
= current
->signal
->cstime
;
1155 spin_unlock_irq(¤t
->sighand
->siglock
);
1156 tms
->tms_utime
= cputime_to_clock_t(tgutime
);
1157 tms
->tms_stime
= cputime_to_clock_t(tgstime
);
1158 tms
->tms_cutime
= cputime_to_clock_t(cutime
);
1159 tms
->tms_cstime
= cputime_to_clock_t(cstime
);
1162 SYSCALL_DEFINE1(times
, struct tms __user
*, tbuf
)
1168 if (copy_to_user(tbuf
, &tmp
, sizeof(struct tms
)))
1171 force_successful_syscall_return();
1172 return (long) jiffies_64_to_clock_t(get_jiffies_64());
1176 * This needs some heavy checking ...
1177 * I just haven't the stomach for it. I also don't fully
1178 * understand sessions/pgrp etc. Let somebody who does explain it.
1180 * OK, I think I have the protection semantics right.... this is really
1181 * only important on a multi-user system anyway, to make sure one user
1182 * can't send a signal to a process owned by another. -TYT, 12/12/91
1184 * Auch. Had to add the 'did_exec' flag to conform completely to POSIX.
1187 SYSCALL_DEFINE2(setpgid
, pid_t
, pid
, pid_t
, pgid
)
1189 struct task_struct
*p
;
1190 struct task_struct
*group_leader
= current
->group_leader
;
1195 pid
= task_pid_vnr(group_leader
);
1202 /* From this point forward we keep holding onto the tasklist lock
1203 * so that our parent does not change from under us. -DaveM
1205 write_lock_irq(&tasklist_lock
);
1208 p
= find_task_by_vpid(pid
);
1213 if (!thread_group_leader(p
))
1216 if (same_thread_group(p
->real_parent
, group_leader
)) {
1218 if (task_session(p
) != task_session(group_leader
))
1225 if (p
!= group_leader
)
1230 if (p
->signal
->leader
)
1235 struct task_struct
*g
;
1237 pgrp
= find_vpid(pgid
);
1238 g
= pid_task(pgrp
, PIDTYPE_PGID
);
1239 if (!g
|| task_session(g
) != task_session(group_leader
))
1243 err
= security_task_setpgid(p
, pgid
);
1247 if (task_pgrp(p
) != pgrp
)
1248 change_pid(p
, PIDTYPE_PGID
, pgrp
);
1252 /* All paths lead to here, thus we are safe. -DaveM */
1253 write_unlock_irq(&tasklist_lock
);
1258 SYSCALL_DEFINE1(getpgid
, pid_t
, pid
)
1260 struct task_struct
*p
;
1266 grp
= task_pgrp(current
);
1269 p
= find_task_by_vpid(pid
);
1276 retval
= security_task_getpgid(p
);
1280 retval
= pid_vnr(grp
);
1286 #ifdef __ARCH_WANT_SYS_GETPGRP
1288 SYSCALL_DEFINE0(getpgrp
)
1290 return sys_getpgid(0);
1295 SYSCALL_DEFINE1(getsid
, pid_t
, pid
)
1297 struct task_struct
*p
;
1303 sid
= task_session(current
);
1306 p
= find_task_by_vpid(pid
);
1309 sid
= task_session(p
);
1313 retval
= security_task_getsid(p
);
1317 retval
= pid_vnr(sid
);
1323 SYSCALL_DEFINE0(setsid
)
1325 struct task_struct
*group_leader
= current
->group_leader
;
1326 struct pid
*sid
= task_pid(group_leader
);
1327 pid_t session
= pid_vnr(sid
);
1330 write_lock_irq(&tasklist_lock
);
1331 /* Fail if I am already a session leader */
1332 if (group_leader
->signal
->leader
)
1335 /* Fail if a process group id already exists that equals the
1336 * proposed session id.
1338 if (pid_task(sid
, PIDTYPE_PGID
))
1341 group_leader
->signal
->leader
= 1;
1342 __set_special_pids(sid
);
1344 proc_clear_tty(group_leader
);
1348 write_unlock_irq(&tasklist_lock
);
1350 proc_sid_connector(group_leader
);
1351 sched_autogroup_create_attach(group_leader
);
1356 DECLARE_RWSEM(uts_sem
);
1358 #ifdef COMPAT_UTS_MACHINE
1359 #define override_architecture(name) \
1360 (personality(current->personality) == PER_LINUX32 && \
1361 copy_to_user(name->machine, COMPAT_UTS_MACHINE, \
1362 sizeof(COMPAT_UTS_MACHINE)))
1364 #define override_architecture(name) 0
1368 * Work around broken programs that cannot handle "Linux 3.0".
1369 * Instead we map 3.x to 2.6.40+x, so e.g. 3.0 would be 2.6.40
1371 static int override_release(char __user
*release
, size_t len
)
1375 if (current
->personality
& UNAME26
) {
1376 const char *rest
= UTS_RELEASE
;
1377 char buf
[65] = { 0 };
1383 if (*rest
== '.' && ++ndots
>= 3)
1385 if (!isdigit(*rest
) && *rest
!= '.')
1389 v
= ((LINUX_VERSION_CODE
>> 8) & 0xff) + 40;
1390 copy
= clamp_t(size_t, len
, 1, sizeof(buf
));
1391 copy
= scnprintf(buf
, copy
, "2.6.%u%s", v
, rest
);
1392 ret
= copy_to_user(release
, buf
, copy
+ 1);
1397 SYSCALL_DEFINE1(newuname
, struct new_utsname __user
*, name
)
1401 down_read(&uts_sem
);
1402 if (copy_to_user(name
, utsname(), sizeof *name
))
1406 if (!errno
&& override_release(name
->release
, sizeof(name
->release
)))
1408 if (!errno
&& override_architecture(name
))
1413 #ifdef __ARCH_WANT_SYS_OLD_UNAME
1417 SYSCALL_DEFINE1(uname
, struct old_utsname __user
*, name
)
1424 down_read(&uts_sem
);
1425 if (copy_to_user(name
, utsname(), sizeof(*name
)))
1429 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1431 if (!error
&& override_architecture(name
))
1436 SYSCALL_DEFINE1(olduname
, struct oldold_utsname __user
*, name
)
1442 if (!access_ok(VERIFY_WRITE
, name
, sizeof(struct oldold_utsname
)))
1445 down_read(&uts_sem
);
1446 error
= __copy_to_user(&name
->sysname
, &utsname()->sysname
,
1448 error
|= __put_user(0, name
->sysname
+ __OLD_UTS_LEN
);
1449 error
|= __copy_to_user(&name
->nodename
, &utsname()->nodename
,
1451 error
|= __put_user(0, name
->nodename
+ __OLD_UTS_LEN
);
1452 error
|= __copy_to_user(&name
->release
, &utsname()->release
,
1454 error
|= __put_user(0, name
->release
+ __OLD_UTS_LEN
);
1455 error
|= __copy_to_user(&name
->version
, &utsname()->version
,
1457 error
|= __put_user(0, name
->version
+ __OLD_UTS_LEN
);
1458 error
|= __copy_to_user(&name
->machine
, &utsname()->machine
,
1460 error
|= __put_user(0, name
->machine
+ __OLD_UTS_LEN
);
1463 if (!error
&& override_architecture(name
))
1465 if (!error
&& override_release(name
->release
, sizeof(name
->release
)))
1467 return error
? -EFAULT
: 0;
1471 SYSCALL_DEFINE2(sethostname
, char __user
*, name
, int, len
)
1474 char tmp
[__NEW_UTS_LEN
];
1476 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1479 if (len
< 0 || len
> __NEW_UTS_LEN
)
1481 down_write(&uts_sem
);
1483 if (!copy_from_user(tmp
, name
, len
)) {
1484 struct new_utsname
*u
= utsname();
1486 memcpy(u
->nodename
, tmp
, len
);
1487 memset(u
->nodename
+ len
, 0, sizeof(u
->nodename
) - len
);
1489 uts_proc_notify(UTS_PROC_HOSTNAME
);
1495 #ifdef __ARCH_WANT_SYS_GETHOSTNAME
1497 SYSCALL_DEFINE2(gethostname
, char __user
*, name
, int, len
)
1500 struct new_utsname
*u
;
1504 down_read(&uts_sem
);
1506 i
= 1 + strlen(u
->nodename
);
1510 if (copy_to_user(name
, u
->nodename
, i
))
1519 * Only setdomainname; getdomainname can be implemented by calling
1522 SYSCALL_DEFINE2(setdomainname
, char __user
*, name
, int, len
)
1525 char tmp
[__NEW_UTS_LEN
];
1527 if (!ns_capable(current
->nsproxy
->uts_ns
->user_ns
, CAP_SYS_ADMIN
))
1529 if (len
< 0 || len
> __NEW_UTS_LEN
)
1532 down_write(&uts_sem
);
1534 if (!copy_from_user(tmp
, name
, len
)) {
1535 struct new_utsname
*u
= utsname();
1537 memcpy(u
->domainname
, tmp
, len
);
1538 memset(u
->domainname
+ len
, 0, sizeof(u
->domainname
) - len
);
1540 uts_proc_notify(UTS_PROC_DOMAINNAME
);
1546 SYSCALL_DEFINE2(getrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1548 struct rlimit value
;
1551 ret
= do_prlimit(current
, resource
, NULL
, &value
);
1553 ret
= copy_to_user(rlim
, &value
, sizeof(*rlim
)) ? -EFAULT
: 0;
1558 #ifdef __ARCH_WANT_SYS_OLD_GETRLIMIT
1561 * Back compatibility for getrlimit. Needed for some apps.
1564 SYSCALL_DEFINE2(old_getrlimit
, unsigned int, resource
,
1565 struct rlimit __user
*, rlim
)
1568 if (resource
>= RLIM_NLIMITS
)
1571 task_lock(current
->group_leader
);
1572 x
= current
->signal
->rlim
[resource
];
1573 task_unlock(current
->group_leader
);
1574 if (x
.rlim_cur
> 0x7FFFFFFF)
1575 x
.rlim_cur
= 0x7FFFFFFF;
1576 if (x
.rlim_max
> 0x7FFFFFFF)
1577 x
.rlim_max
= 0x7FFFFFFF;
1578 return copy_to_user(rlim
, &x
, sizeof(x
))?-EFAULT
:0;
1583 static inline bool rlim64_is_infinity(__u64 rlim64
)
1585 #if BITS_PER_LONG < 64
1586 return rlim64
>= ULONG_MAX
;
1588 return rlim64
== RLIM64_INFINITY
;
1592 static void rlim_to_rlim64(const struct rlimit
*rlim
, struct rlimit64
*rlim64
)
1594 if (rlim
->rlim_cur
== RLIM_INFINITY
)
1595 rlim64
->rlim_cur
= RLIM64_INFINITY
;
1597 rlim64
->rlim_cur
= rlim
->rlim_cur
;
1598 if (rlim
->rlim_max
== RLIM_INFINITY
)
1599 rlim64
->rlim_max
= RLIM64_INFINITY
;
1601 rlim64
->rlim_max
= rlim
->rlim_max
;
1604 static void rlim64_to_rlim(const struct rlimit64
*rlim64
, struct rlimit
*rlim
)
1606 if (rlim64_is_infinity(rlim64
->rlim_cur
))
1607 rlim
->rlim_cur
= RLIM_INFINITY
;
1609 rlim
->rlim_cur
= (unsigned long)rlim64
->rlim_cur
;
1610 if (rlim64_is_infinity(rlim64
->rlim_max
))
1611 rlim
->rlim_max
= RLIM_INFINITY
;
1613 rlim
->rlim_max
= (unsigned long)rlim64
->rlim_max
;
1616 /* make sure you are allowed to change @tsk limits before calling this */
1617 int do_prlimit(struct task_struct
*tsk
, unsigned int resource
,
1618 struct rlimit
*new_rlim
, struct rlimit
*old_rlim
)
1620 struct rlimit
*rlim
;
1623 if (resource
>= RLIM_NLIMITS
)
1626 if (new_rlim
->rlim_cur
> new_rlim
->rlim_max
)
1628 if (resource
== RLIMIT_NOFILE
&&
1629 new_rlim
->rlim_max
> sysctl_nr_open
)
1633 /* protect tsk->signal and tsk->sighand from disappearing */
1634 read_lock(&tasklist_lock
);
1635 if (!tsk
->sighand
) {
1640 rlim
= tsk
->signal
->rlim
+ resource
;
1641 task_lock(tsk
->group_leader
);
1643 /* Keep the capable check against init_user_ns until
1644 cgroups can contain all limits */
1645 if (new_rlim
->rlim_max
> rlim
->rlim_max
&&
1646 !capable(CAP_SYS_RESOURCE
))
1649 retval
= security_task_setrlimit(tsk
->group_leader
,
1650 resource
, new_rlim
);
1651 if (resource
== RLIMIT_CPU
&& new_rlim
->rlim_cur
== 0) {
1653 * The caller is asking for an immediate RLIMIT_CPU
1654 * expiry. But we use the zero value to mean "it was
1655 * never set". So let's cheat and make it one second
1658 new_rlim
->rlim_cur
= 1;
1667 task_unlock(tsk
->group_leader
);
1670 * RLIMIT_CPU handling. Note that the kernel fails to return an error
1671 * code if it rejected the user's attempt to set RLIMIT_CPU. This is a
1672 * very long-standing error, and fixing it now risks breakage of
1673 * applications, so we live with it
1675 if (!retval
&& new_rlim
&& resource
== RLIMIT_CPU
&&
1676 new_rlim
->rlim_cur
!= RLIM_INFINITY
)
1677 update_rlimit_cpu(tsk
, new_rlim
->rlim_cur
);
1679 read_unlock(&tasklist_lock
);
1683 /* rcu lock must be held */
1684 static int check_prlimit_permission(struct task_struct
*task
)
1686 const struct cred
*cred
= current_cred(), *tcred
;
1688 if (current
== task
)
1691 tcred
= __task_cred(task
);
1692 if (uid_eq(cred
->uid
, tcred
->euid
) &&
1693 uid_eq(cred
->uid
, tcred
->suid
) &&
1694 uid_eq(cred
->uid
, tcred
->uid
) &&
1695 gid_eq(cred
->gid
, tcred
->egid
) &&
1696 gid_eq(cred
->gid
, tcred
->sgid
) &&
1697 gid_eq(cred
->gid
, tcred
->gid
))
1699 if (ns_capable(tcred
->user_ns
, CAP_SYS_RESOURCE
))
1705 SYSCALL_DEFINE4(prlimit64
, pid_t
, pid
, unsigned int, resource
,
1706 const struct rlimit64 __user
*, new_rlim
,
1707 struct rlimit64 __user
*, old_rlim
)
1709 struct rlimit64 old64
, new64
;
1710 struct rlimit old
, new;
1711 struct task_struct
*tsk
;
1715 if (copy_from_user(&new64
, new_rlim
, sizeof(new64
)))
1717 rlim64_to_rlim(&new64
, &new);
1721 tsk
= pid
? find_task_by_vpid(pid
) : current
;
1726 ret
= check_prlimit_permission(tsk
);
1731 get_task_struct(tsk
);
1734 ret
= do_prlimit(tsk
, resource
, new_rlim
? &new : NULL
,
1735 old_rlim
? &old
: NULL
);
1737 if (!ret
&& old_rlim
) {
1738 rlim_to_rlim64(&old
, &old64
);
1739 if (copy_to_user(old_rlim
, &old64
, sizeof(old64
)))
1743 put_task_struct(tsk
);
1747 SYSCALL_DEFINE2(setrlimit
, unsigned int, resource
, struct rlimit __user
*, rlim
)
1749 struct rlimit new_rlim
;
1751 if (copy_from_user(&new_rlim
, rlim
, sizeof(*rlim
)))
1753 return do_prlimit(current
, resource
, &new_rlim
, NULL
);
1757 * It would make sense to put struct rusage in the task_struct,
1758 * except that would make the task_struct be *really big*. After
1759 * task_struct gets moved into malloc'ed memory, it would
1760 * make sense to do this. It will make moving the rest of the information
1761 * a lot simpler! (Which we're not doing right now because we're not
1762 * measuring them yet).
1764 * When sampling multiple threads for RUSAGE_SELF, under SMP we might have
1765 * races with threads incrementing their own counters. But since word
1766 * reads are atomic, we either get new values or old values and we don't
1767 * care which for the sums. We always take the siglock to protect reading
1768 * the c* fields from p->signal from races with exit.c updating those
1769 * fields when reaping, so a sample either gets all the additions of a
1770 * given child after it's reaped, or none so this sample is before reaping.
1773 * We need to take the siglock for CHILDEREN, SELF and BOTH
1774 * for the cases current multithreaded, non-current single threaded
1775 * non-current multithreaded. Thread traversal is now safe with
1777 * Strictly speaking, we donot need to take the siglock if we are current and
1778 * single threaded, as no one else can take our signal_struct away, no one
1779 * else can reap the children to update signal->c* counters, and no one else
1780 * can race with the signal-> fields. If we do not take any lock, the
1781 * signal-> fields could be read out of order while another thread was just
1782 * exiting. So we should place a read memory barrier when we avoid the lock.
1783 * On the writer side, write memory barrier is implied in __exit_signal
1784 * as __exit_signal releases the siglock spinlock after updating the signal->
1785 * fields. But we don't do this yet to keep things simple.
1789 static void accumulate_thread_rusage(struct task_struct
*t
, struct rusage
*r
)
1791 r
->ru_nvcsw
+= t
->nvcsw
;
1792 r
->ru_nivcsw
+= t
->nivcsw
;
1793 r
->ru_minflt
+= t
->min_flt
;
1794 r
->ru_majflt
+= t
->maj_flt
;
1795 r
->ru_inblock
+= task_io_get_inblock(t
);
1796 r
->ru_oublock
+= task_io_get_oublock(t
);
1799 static void k_getrusage(struct task_struct
*p
, int who
, struct rusage
*r
)
1801 struct task_struct
*t
;
1802 unsigned long flags
;
1803 cputime_t tgutime
, tgstime
, utime
, stime
;
1804 unsigned long maxrss
= 0;
1806 memset((char *) r
, 0, sizeof *r
);
1809 if (who
== RUSAGE_THREAD
) {
1810 task_cputime_adjusted(current
, &utime
, &stime
);
1811 accumulate_thread_rusage(p
, r
);
1812 maxrss
= p
->signal
->maxrss
;
1816 if (!lock_task_sighand(p
, &flags
))
1821 case RUSAGE_CHILDREN
:
1822 utime
= p
->signal
->cutime
;
1823 stime
= p
->signal
->cstime
;
1824 r
->ru_nvcsw
= p
->signal
->cnvcsw
;
1825 r
->ru_nivcsw
= p
->signal
->cnivcsw
;
1826 r
->ru_minflt
= p
->signal
->cmin_flt
;
1827 r
->ru_majflt
= p
->signal
->cmaj_flt
;
1828 r
->ru_inblock
= p
->signal
->cinblock
;
1829 r
->ru_oublock
= p
->signal
->coublock
;
1830 maxrss
= p
->signal
->cmaxrss
;
1832 if (who
== RUSAGE_CHILDREN
)
1836 thread_group_cputime_adjusted(p
, &tgutime
, &tgstime
);
1839 r
->ru_nvcsw
+= p
->signal
->nvcsw
;
1840 r
->ru_nivcsw
+= p
->signal
->nivcsw
;
1841 r
->ru_minflt
+= p
->signal
->min_flt
;
1842 r
->ru_majflt
+= p
->signal
->maj_flt
;
1843 r
->ru_inblock
+= p
->signal
->inblock
;
1844 r
->ru_oublock
+= p
->signal
->oublock
;
1845 if (maxrss
< p
->signal
->maxrss
)
1846 maxrss
= p
->signal
->maxrss
;
1849 accumulate_thread_rusage(t
, r
);
1857 unlock_task_sighand(p
, &flags
);
1860 cputime_to_timeval(utime
, &r
->ru_utime
);
1861 cputime_to_timeval(stime
, &r
->ru_stime
);
1863 if (who
!= RUSAGE_CHILDREN
) {
1864 struct mm_struct
*mm
= get_task_mm(p
);
1866 setmax_mm_hiwater_rss(&maxrss
, mm
);
1870 r
->ru_maxrss
= maxrss
* (PAGE_SIZE
/ 1024); /* convert pages to KBs */
1873 int getrusage(struct task_struct
*p
, int who
, struct rusage __user
*ru
)
1876 k_getrusage(p
, who
, &r
);
1877 return copy_to_user(ru
, &r
, sizeof(r
)) ? -EFAULT
: 0;
1880 SYSCALL_DEFINE2(getrusage
, int, who
, struct rusage __user
*, ru
)
1882 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1883 who
!= RUSAGE_THREAD
)
1885 return getrusage(current
, who
, ru
);
1888 #ifdef CONFIG_COMPAT
1889 COMPAT_SYSCALL_DEFINE2(getrusage
, int, who
, struct compat_rusage __user
*, ru
)
1893 if (who
!= RUSAGE_SELF
&& who
!= RUSAGE_CHILDREN
&&
1894 who
!= RUSAGE_THREAD
)
1897 k_getrusage(current
, who
, &r
);
1898 return put_compat_rusage(&r
, ru
);
1902 SYSCALL_DEFINE1(umask
, int, mask
)
1904 mask
= xchg(¤t
->fs
->umask
, mask
& S_IRWXUGO
);
1908 static int prctl_set_mm_exe_file(struct mm_struct
*mm
, unsigned int fd
)
1911 struct inode
*inode
;
1918 inode
= file_inode(exe
.file
);
1921 * Because the original mm->exe_file points to executable file, make
1922 * sure that this one is executable as well, to avoid breaking an
1926 if (!S_ISREG(inode
->i_mode
) ||
1927 exe
.file
->f_path
.mnt
->mnt_flags
& MNT_NOEXEC
)
1930 err
= inode_permission(inode
, MAY_EXEC
);
1934 down_write(&mm
->mmap_sem
);
1937 * Forbid mm->exe_file change if old file still mapped.
1941 struct vm_area_struct
*vma
;
1943 for (vma
= mm
->mmap
; vma
; vma
= vma
->vm_next
)
1945 path_equal(&vma
->vm_file
->f_path
,
1946 &mm
->exe_file
->f_path
))
1951 * The symlink can be changed only once, just to disallow arbitrary
1952 * transitions malicious software might bring in. This means one
1953 * could make a snapshot over all processes running and monitor
1954 * /proc/pid/exe changes to notice unusual activity if needed.
1957 if (test_and_set_bit(MMF_EXE_FILE_CHANGED
, &mm
->flags
))
1961 set_mm_exe_file(mm
, exe
.file
); /* this grabs a reference to exe.file */
1963 up_write(&mm
->mmap_sem
);
1970 static int prctl_set_mm(int opt
, unsigned long addr
,
1971 unsigned long arg4
, unsigned long arg5
)
1973 unsigned long rlim
= rlimit(RLIMIT_DATA
);
1974 struct mm_struct
*mm
= current
->mm
;
1975 struct vm_area_struct
*vma
;
1978 if (arg5
|| (arg4
&& opt
!= PR_SET_MM_AUXV
))
1981 if (!capable(CAP_SYS_RESOURCE
))
1984 if (opt
== PR_SET_MM_EXE_FILE
)
1985 return prctl_set_mm_exe_file(mm
, (unsigned int)addr
);
1987 if (addr
>= TASK_SIZE
|| addr
< mmap_min_addr
)
1992 down_read(&mm
->mmap_sem
);
1993 vma
= find_vma(mm
, addr
);
1996 case PR_SET_MM_START_CODE
:
1997 mm
->start_code
= addr
;
1999 case PR_SET_MM_END_CODE
:
2000 mm
->end_code
= addr
;
2002 case PR_SET_MM_START_DATA
:
2003 mm
->start_data
= addr
;
2005 case PR_SET_MM_END_DATA
:
2006 mm
->end_data
= addr
;
2009 case PR_SET_MM_START_BRK
:
2010 if (addr
<= mm
->end_data
)
2013 if (rlim
< RLIM_INFINITY
&&
2015 (mm
->end_data
- mm
->start_data
) > rlim
)
2018 mm
->start_brk
= addr
;
2022 if (addr
<= mm
->end_data
)
2025 if (rlim
< RLIM_INFINITY
&&
2026 (addr
- mm
->start_brk
) +
2027 (mm
->end_data
- mm
->start_data
) > rlim
)
2034 * If command line arguments and environment
2035 * are placed somewhere else on stack, we can
2036 * set them up here, ARG_START/END to setup
2037 * command line argumets and ENV_START/END
2040 case PR_SET_MM_START_STACK
:
2041 case PR_SET_MM_ARG_START
:
2042 case PR_SET_MM_ARG_END
:
2043 case PR_SET_MM_ENV_START
:
2044 case PR_SET_MM_ENV_END
:
2049 if (opt
== PR_SET_MM_START_STACK
)
2050 mm
->start_stack
= addr
;
2051 else if (opt
== PR_SET_MM_ARG_START
)
2052 mm
->arg_start
= addr
;
2053 else if (opt
== PR_SET_MM_ARG_END
)
2055 else if (opt
== PR_SET_MM_ENV_START
)
2056 mm
->env_start
= addr
;
2057 else if (opt
== PR_SET_MM_ENV_END
)
2062 * This doesn't move auxiliary vector itself
2063 * since it's pinned to mm_struct, but allow
2064 * to fill vector with new values. It's up
2065 * to a caller to provide sane values here
2066 * otherwise user space tools which use this
2067 * vector might be unhappy.
2069 case PR_SET_MM_AUXV
: {
2070 unsigned long user_auxv
[AT_VECTOR_SIZE
];
2072 if (arg4
> sizeof(user_auxv
))
2074 up_read(&mm
->mmap_sem
);
2076 if (copy_from_user(user_auxv
, (const void __user
*)addr
, arg4
))
2079 /* Make sure the last entry is always AT_NULL */
2080 user_auxv
[AT_VECTOR_SIZE
- 2] = 0;
2081 user_auxv
[AT_VECTOR_SIZE
- 1] = 0;
2083 BUILD_BUG_ON(sizeof(user_auxv
) != sizeof(mm
->saved_auxv
));
2086 memcpy(mm
->saved_auxv
, user_auxv
, arg4
);
2087 task_unlock(current
);
2097 up_read(&mm
->mmap_sem
);
2101 #ifdef CONFIG_CHECKPOINT_RESTORE
2102 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2104 return put_user(me
->clear_child_tid
, tid_addr
);
2107 static int prctl_get_tid_address(struct task_struct
*me
, int __user
**tid_addr
)
2114 static int prctl_update_vma_anon_name(struct vm_area_struct
*vma
,
2115 struct vm_area_struct
**prev
,
2116 unsigned long start
, unsigned long end
,
2117 const char __user
*name_addr
)
2119 struct mm_struct
* mm
= vma
->vm_mm
;
2123 if (name_addr
== vma_get_anon_name(vma
)) {
2128 pgoff
= vma
->vm_pgoff
+ ((start
- vma
->vm_start
) >> PAGE_SHIFT
);
2129 *prev
= vma_merge(mm
, *prev
, start
, end
, vma
->vm_flags
, vma
->anon_vma
,
2130 vma
->vm_file
, pgoff
, vma_policy(vma
),
2139 if (start
!= vma
->vm_start
) {
2140 error
= split_vma(mm
, vma
, start
, 1);
2145 if (end
!= vma
->vm_end
) {
2146 error
= split_vma(mm
, vma
, end
, 0);
2153 vma
->shared
.anon_name
= name_addr
;
2156 if (error
== -ENOMEM
)
2161 static int prctl_set_vma_anon_name(unsigned long start
, unsigned long end
,
2165 struct vm_area_struct
* vma
, *prev
;
2166 int unmapped_error
= 0;
2167 int error
= -EINVAL
;
2170 * If the interval [start,end) covers some unmapped address
2171 * ranges, just ignore them, but return -ENOMEM at the end.
2172 * - this matches the handling in madvise.
2174 vma
= find_vma_prev(current
->mm
, start
, &prev
);
2175 if (vma
&& start
> vma
->vm_start
)
2179 /* Still start < end. */
2184 /* Here start < (end|vma->vm_end). */
2185 if (start
< vma
->vm_start
) {
2186 unmapped_error
= -ENOMEM
;
2187 start
= vma
->vm_start
;
2192 /* Here vma->vm_start <= start < (end|vma->vm_end) */
2197 /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */
2198 error
= prctl_update_vma_anon_name(vma
, &prev
, start
, tmp
,
2199 (const char __user
*)arg
);
2203 if (prev
&& start
< prev
->vm_end
)
2204 start
= prev
->vm_end
;
2205 error
= unmapped_error
;
2209 vma
= prev
->vm_next
;
2210 else /* madvise_remove dropped mmap_sem */
2211 vma
= find_vma(current
->mm
, start
);
2215 static int prctl_set_vma(unsigned long opt
, unsigned long start
,
2216 unsigned long len_in
, unsigned long arg
)
2218 struct mm_struct
*mm
= current
->mm
;
2223 #ifndef CONFIG_MT_ENG_BUILD
2224 /* Do not do prctl_set_vma in !eng load */
2228 if (start
& ~PAGE_MASK
)
2230 len
= (len_in
+ ~PAGE_MASK
) & PAGE_MASK
;
2232 /* Check to see whether len was rounded up from small -ve to zero */
2243 down_write(&mm
->mmap_sem
);
2246 case PR_SET_VMA_ANON_NAME
:
2247 error
= prctl_set_vma_anon_name(start
, end
, arg
);
2253 up_write(&mm
->mmap_sem
);
2257 #else /* CONFIG_MMU */
2258 static int prctl_set_vma(unsigned long opt
, unsigned long start
,
2259 unsigned long len_in
, unsigned long arg
)
2265 SYSCALL_DEFINE5(prctl
, int, option
, unsigned long, arg2
, unsigned long, arg3
,
2266 unsigned long, arg4
, unsigned long, arg5
)
2268 struct task_struct
*me
= current
;
2269 struct task_struct
*tsk
;
2270 unsigned char comm
[sizeof(me
->comm
)];
2273 error
= security_task_prctl(option
, arg2
, arg3
, arg4
, arg5
);
2274 if (error
!= -ENOSYS
)
2279 case PR_SET_PDEATHSIG
:
2280 if (!valid_signal(arg2
)) {
2284 me
->pdeath_signal
= arg2
;
2286 case PR_GET_PDEATHSIG
:
2287 error
= put_user(me
->pdeath_signal
, (int __user
*)arg2
);
2289 case PR_GET_DUMPABLE
:
2290 error
= get_dumpable(me
->mm
);
2292 case PR_SET_DUMPABLE
:
2293 if (arg2
!= SUID_DUMP_DISABLE
&& arg2
!= SUID_DUMP_USER
) {
2297 set_dumpable(me
->mm
, arg2
);
2300 case PR_SET_UNALIGN
:
2301 error
= SET_UNALIGN_CTL(me
, arg2
);
2303 case PR_GET_UNALIGN
:
2304 error
= GET_UNALIGN_CTL(me
, arg2
);
2307 error
= SET_FPEMU_CTL(me
, arg2
);
2310 error
= GET_FPEMU_CTL(me
, arg2
);
2313 error
= SET_FPEXC_CTL(me
, arg2
);
2316 error
= GET_FPEXC_CTL(me
, arg2
);
2319 error
= PR_TIMING_STATISTICAL
;
2322 if (arg2
!= PR_TIMING_STATISTICAL
)
2326 comm
[sizeof(me
->comm
) - 1] = 0;
2327 if (strncpy_from_user(comm
, (char __user
*)arg2
,
2328 sizeof(me
->comm
) - 1) < 0)
2330 set_task_comm(me
, comm
);
2331 proc_comm_connector(me
);
2334 get_task_comm(comm
, me
);
2335 if (copy_to_user((char __user
*)arg2
, comm
, sizeof(comm
)))
2339 error
= GET_ENDIAN(me
, arg2
);
2342 error
= SET_ENDIAN(me
, arg2
);
2344 case PR_GET_SECCOMP
:
2345 error
= prctl_get_seccomp();
2347 case PR_SET_SECCOMP
:
2348 error
= prctl_set_seccomp(arg2
, (char __user
*)arg3
);
2351 error
= GET_TSC_CTL(arg2
);
2354 error
= SET_TSC_CTL(arg2
);
2356 case PR_TASK_PERF_EVENTS_DISABLE
:
2357 error
= perf_event_task_disable();
2359 case PR_TASK_PERF_EVENTS_ENABLE
:
2360 error
= perf_event_task_enable();
2362 case PR_GET_TIMERSLACK
:
2363 error
= current
->timer_slack_ns
;
2365 case PR_SET_TIMERSLACK
:
2367 current
->timer_slack_ns
=
2368 current
->default_timer_slack_ns
;
2370 current
->timer_slack_ns
= arg2
;
2376 case PR_MCE_KILL_CLEAR
:
2379 current
->flags
&= ~PF_MCE_PROCESS
;
2381 case PR_MCE_KILL_SET
:
2382 current
->flags
|= PF_MCE_PROCESS
;
2383 if (arg3
== PR_MCE_KILL_EARLY
)
2384 current
->flags
|= PF_MCE_EARLY
;
2385 else if (arg3
== PR_MCE_KILL_LATE
)
2386 current
->flags
&= ~PF_MCE_EARLY
;
2387 else if (arg3
== PR_MCE_KILL_DEFAULT
)
2389 ~(PF_MCE_EARLY
|PF_MCE_PROCESS
);
2393 case PR_SET_TIMERSLACK_PID
:
2394 if (current
->pid
!= (pid_t
)arg3
&&
2395 !capable(CAP_SYS_NICE
))
2398 tsk
= find_task_by_pid_ns((pid_t
)arg3
, &init_pid_ns
);
2403 get_task_struct(tsk
);
2406 tsk
->timer_slack_ns
=
2407 tsk
->default_timer_slack_ns
;
2409 tsk
->timer_slack_ns
= arg2
;
2410 put_task_struct(tsk
);
2417 case PR_MCE_KILL_GET
:
2418 if (arg2
| arg3
| arg4
| arg5
)
2420 if (current
->flags
& PF_MCE_PROCESS
)
2421 error
= (current
->flags
& PF_MCE_EARLY
) ?
2422 PR_MCE_KILL_EARLY
: PR_MCE_KILL_LATE
;
2424 error
= PR_MCE_KILL_DEFAULT
;
2427 error
= prctl_set_mm(arg2
, arg3
, arg4
, arg5
);
2429 case PR_GET_TID_ADDRESS
:
2430 error
= prctl_get_tid_address(me
, (int __user
**)arg2
);
2432 case PR_SET_CHILD_SUBREAPER
:
2433 me
->signal
->is_child_subreaper
= !!arg2
;
2435 case PR_GET_CHILD_SUBREAPER
:
2436 error
= put_user(me
->signal
->is_child_subreaper
,
2437 (int __user
*)arg2
);
2439 case PR_SET_NO_NEW_PRIVS
:
2440 if (arg2
!= 1 || arg3
|| arg4
|| arg5
)
2443 task_set_no_new_privs(current
);
2445 case PR_GET_NO_NEW_PRIVS
:
2446 if (arg2
|| arg3
|| arg4
|| arg5
)
2448 return task_no_new_privs(current
) ? 1 : 0;
2450 error
= prctl_set_vma(arg2
, arg3
, arg4
, arg5
);
2459 SYSCALL_DEFINE3(getcpu
, unsigned __user
*, cpup
, unsigned __user
*, nodep
,
2460 struct getcpu_cache __user
*, unused
)
2463 int cpu
= raw_smp_processor_id();
2465 err
|= put_user(cpu
, cpup
);
2467 err
|= put_user(cpu_to_node(cpu
), nodep
);
2468 return err
? -EFAULT
: 0;
2471 char poweroff_cmd
[POWEROFF_CMD_PATH_LEN
] = "/sbin/poweroff";
2473 static int __orderly_poweroff(bool force
)
2476 static char *envp
[] = {
2478 "PATH=/sbin:/bin:/usr/sbin:/usr/bin",
2483 argv
= argv_split(GFP_KERNEL
, poweroff_cmd
, NULL
);
2485 ret
= call_usermodehelper(argv
[0], argv
, envp
, UMH_WAIT_EXEC
);
2488 printk(KERN_WARNING
"%s failed to allocate memory for \"%s\"\n",
2489 __func__
, poweroff_cmd
);
2494 printk(KERN_WARNING
"Failed to start orderly shutdown: "
2495 "forcing the issue\n");
2497 * I guess this should try to kick off some daemon to sync and
2498 * poweroff asap. Or not even bother syncing if we're doing an
2499 * emergency shutdown?
2508 static bool poweroff_force
;
2510 static void poweroff_work_func(struct work_struct
*work
)
2512 __orderly_poweroff(poweroff_force
);
2515 static DECLARE_WORK(poweroff_work
, poweroff_work_func
);
2518 * orderly_poweroff - Trigger an orderly system poweroff
2519 * @force: force poweroff if command execution fails
2521 * This may be called from any context to trigger a system shutdown.
2522 * If the orderly shutdown fails, it will force an immediate shutdown.
2524 int orderly_poweroff(bool force
)
2526 if (force
) /* do not override the pending "true" */
2527 poweroff_force
= true;
2528 schedule_work(&poweroff_work
);
2531 EXPORT_SYMBOL_GPL(orderly_poweroff
);
2534 * do_sysinfo - fill in sysinfo struct
2535 * @info: pointer to buffer to fill
2537 static int do_sysinfo(struct sysinfo
*info
)
2539 unsigned long mem_total
, sav_total
;
2540 unsigned int mem_unit
, bitcount
;
2543 memset(info
, 0, sizeof(struct sysinfo
));
2546 monotonic_to_bootbased(&tp
);
2547 info
->uptime
= tp
.tv_sec
+ (tp
.tv_nsec
? 1 : 0);
2549 get_avenrun(info
->loads
, 0, SI_LOAD_SHIFT
- FSHIFT
);
2551 info
->procs
= nr_threads
;
2557 * If the sum of all the available memory (i.e. ram + swap)
2558 * is less than can be stored in a 32 bit unsigned long then
2559 * we can be binary compatible with 2.2.x kernels. If not,
2560 * well, in that case 2.2.x was broken anyways...
2562 * -Erik Andersen <andersee@debian.org>
2565 mem_total
= info
->totalram
+ info
->totalswap
;
2566 if (mem_total
< info
->totalram
|| mem_total
< info
->totalswap
)
2569 mem_unit
= info
->mem_unit
;
2570 while (mem_unit
> 1) {
2573 sav_total
= mem_total
;
2575 if (mem_total
< sav_total
)
2580 * If mem_total did not overflow, multiply all memory values by
2581 * info->mem_unit and set it to 1. This leaves things compatible
2582 * with 2.2.x, and also retains compatibility with earlier 2.4.x
2587 info
->totalram
<<= bitcount
;
2588 info
->freeram
<<= bitcount
;
2589 info
->sharedram
<<= bitcount
;
2590 info
->bufferram
<<= bitcount
;
2591 info
->totalswap
<<= bitcount
;
2592 info
->freeswap
<<= bitcount
;
2593 info
->totalhigh
<<= bitcount
;
2594 info
->freehigh
<<= bitcount
;
2600 SYSCALL_DEFINE1(sysinfo
, struct sysinfo __user
*, info
)
2606 if (copy_to_user(info
, &val
, sizeof(struct sysinfo
)))
2612 #ifdef CONFIG_COMPAT
2613 struct compat_sysinfo
{
2627 char _f
[20-2*sizeof(u32
)-sizeof(int)];
2630 COMPAT_SYSCALL_DEFINE1(sysinfo
, struct compat_sysinfo __user
*, info
)
2636 /* Check to see if any memory value is too large for 32-bit and scale
2639 if ((s
.totalram
>> 32) || (s
.totalswap
>> 32)) {
2642 while (s
.mem_unit
< PAGE_SIZE
) {
2647 s
.totalram
>>= bitcount
;
2648 s
.freeram
>>= bitcount
;
2649 s
.sharedram
>>= bitcount
;
2650 s
.bufferram
>>= bitcount
;
2651 s
.totalswap
>>= bitcount
;
2652 s
.freeswap
>>= bitcount
;
2653 s
.totalhigh
>>= bitcount
;
2654 s
.freehigh
>>= bitcount
;
2657 if (!access_ok(VERIFY_WRITE
, info
, sizeof(struct compat_sysinfo
)) ||
2658 __put_user(s
.uptime
, &info
->uptime
) ||
2659 __put_user(s
.loads
[0], &info
->loads
[0]) ||
2660 __put_user(s
.loads
[1], &info
->loads
[1]) ||
2661 __put_user(s
.loads
[2], &info
->loads
[2]) ||
2662 __put_user(s
.totalram
, &info
->totalram
) ||
2663 __put_user(s
.freeram
, &info
->freeram
) ||
2664 __put_user(s
.sharedram
, &info
->sharedram
) ||
2665 __put_user(s
.bufferram
, &info
->bufferram
) ||
2666 __put_user(s
.totalswap
, &info
->totalswap
) ||
2667 __put_user(s
.freeswap
, &info
->freeswap
) ||
2668 __put_user(s
.procs
, &info
->procs
) ||
2669 __put_user(s
.totalhigh
, &info
->totalhigh
) ||
2670 __put_user(s
.freehigh
, &info
->freehigh
) ||
2671 __put_user(s
.mem_unit
, &info
->mem_unit
))
2676 #endif /* CONFIG_COMPAT */